Cumulus: a federated electronic health record-based learning system powered by Fast Healthcare Interoperability...

Area of Science:

• Health Informatics
• Data Science
• Public Health

Background:

• Large-scale electronic health record (EHR) data exchange faces significant challenges.
• Standardized data access across healthcare systems is crucial for advancing research and public health initiatives.

Purpose of the Study:

• To develop, deploy, and test an open-source, cloud-hosted application ('listener') for accessing standardized EHR data.
• To overcome barriers in large-scale EHR data exchange using the SMART/HL7 Bulk FHIR API.

Main Methods:

• Developed Cumulus, a scalable, federated model for data sharing and learning.
• Utilized containerized, cloud-hosted software within healthcare organizations' security envelopes.
• Integrated artificial intelligence (AI) and natural language processing (NLP) for unstructured text analysis.

Main Results:

• Cumulus successfully accessed EHR data via Bulk FHIR, streamlining automated processing and sharing across 5 healthcare systems.
• Generated aggregated patient counts for population health studies, with all code available open source.
• A policy restricting data sharing to aggregate information facilitated data sharing agreements.

Conclusions:

• Cumulus addresses EHR data sharing barriers through standard API support, cloud computing, and AI advancements.
• The platform demonstrates potential for scalability to support learning across diverse network configurations and use cases.
• Facilitates provider autonomy and administrative simplicity in data sharing.